1 /*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/
4 This file is part of systemd.
6 Copyright 2010 Lennart Poettering
8 systemd is free software; you can redistribute it and/or modify it
9 under the terms of the GNU Lesser General Public License as published by
10 the Free Software Foundation; either version 2.1 of the License, or
11 (at your option) any later version.
13 systemd is distributed in the hope that it will be useful, but
14 WITHOUT ANY WARRANTY; without even the implied warranty of
15 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 Lesser General Public License for more details.
18 You should have received a copy of the GNU Lesser General Public License
19 along with systemd; If not, see <http://www.gnu.org/licenses/>.
26 #include <sys/types.h>
32 #include <sys/prctl.h>
33 #include <sys/mount.h>
35 #ifdef HAVE_VALGRIND_VALGRIND_H
36 #include <valgrind/valgrind.h>
42 #include "sd-daemon.h"
43 #include "sd-messages.h"
47 #include "load-fragment.h"
50 #include "conf-parser.h"
58 #include "architecture.h"
60 #include "path-util.h"
61 #include "switch-root.h"
62 #include "capability.h"
65 #include "clock-util.h"
67 #include "dbus-manager.h"
68 #include "bus-error.h"
71 #include "mount-setup.h"
72 #include "loopback-setup.h"
73 #include "hostname-setup.h"
74 #include "machine-id-setup.h"
75 #include "selinux-setup.h"
76 #include "ima-setup.h"
77 #include "smack-setup.h"
79 #include "kmod-setup.h"
87 ACTION_DUMP_CONFIGURATION_ITEMS,
89 } arg_action = ACTION_RUN;
90 static char *arg_default_unit = NULL;
91 static SystemdRunningAs arg_running_as = _SYSTEMD_RUNNING_AS_INVALID;
92 static bool arg_dump_core = true;
93 static bool arg_crash_shell = false;
94 static int arg_crash_chvt = -1;
95 static bool arg_confirm_spawn = false;
96 static ShowStatus arg_show_status = _SHOW_STATUS_UNSET;
97 static bool arg_switched_root = false;
98 static int arg_no_pager = -1;
99 static char ***arg_join_controllers = NULL;
100 static ExecOutput arg_default_std_output = EXEC_OUTPUT_JOURNAL;
101 static ExecOutput arg_default_std_error = EXEC_OUTPUT_INHERIT;
102 static usec_t arg_default_restart_usec = DEFAULT_RESTART_USEC;
103 static usec_t arg_default_timeout_start_usec = DEFAULT_TIMEOUT_USEC;
104 static usec_t arg_default_timeout_stop_usec = DEFAULT_TIMEOUT_USEC;
105 static usec_t arg_default_start_limit_interval = DEFAULT_START_LIMIT_INTERVAL;
106 static unsigned arg_default_start_limit_burst = DEFAULT_START_LIMIT_BURST;
107 static usec_t arg_runtime_watchdog = 0;
108 static usec_t arg_shutdown_watchdog = 10 * USEC_PER_MINUTE;
109 static char **arg_default_environment = NULL;
110 static struct rlimit *arg_default_rlimit[_RLIMIT_MAX] = {};
111 static uint64_t arg_capability_bounding_set_drop = 0;
112 static nsec_t arg_timer_slack_nsec = NSEC_INFINITY;
113 static usec_t arg_default_timer_accuracy_usec = 1 * USEC_PER_MINUTE;
114 static Set* arg_syscall_archs = NULL;
115 static FILE* arg_serialization = NULL;
116 static bool arg_default_cpu_accounting = false;
117 static bool arg_default_blockio_accounting = false;
118 static bool arg_default_memory_accounting = false;
120 static void nop_handler(int sig) {}
122 static void pager_open_if_enabled(void) {
124 if (arg_no_pager <= 0)
130 noreturn static void crash(int sig) {
133 /* Pass this on immediately, if this is not PID 1 */
135 else if (!arg_dump_core)
136 log_error("Caught <%s>, not dumping core.", signal_to_string(sig));
138 struct sigaction sa = {
139 .sa_handler = nop_handler,
140 .sa_flags = SA_NOCLDSTOP|SA_RESTART,
144 /* We want to wait for the core process, hence let's enable SIGCHLD */
145 sigaction(SIGCHLD, &sa, NULL);
149 log_error("Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig));
152 struct rlimit rl = {};
154 /* Enable default signal handler for core dump */
156 sa.sa_handler = SIG_DFL;
157 sigaction(sig, &sa, NULL);
159 /* Don't limit the core dump size */
160 rl.rlim_cur = RLIM_INFINITY;
161 rl.rlim_max = RLIM_INFINITY;
162 setrlimit(RLIMIT_CORE, &rl);
164 /* Just to be sure... */
167 /* Raise the signal again */
170 assert_not_reached("We shouldn't be here...");
177 /* Order things nicely. */
178 r = wait_for_terminate(pid, &status);
180 log_error("Caught <%s>, waitpid() failed: %s", signal_to_string(sig), strerror(-r));
181 else if (status.si_code != CLD_DUMPED)
182 log_error("Caught <%s>, core dump failed.", signal_to_string(sig));
184 log_error("Caught <%s>, dumped core as pid "PID_FMT".", signal_to_string(sig), pid);
189 chvt(arg_crash_chvt);
191 if (arg_crash_shell) {
192 struct sigaction sa = {
193 .sa_handler = SIG_IGN,
194 .sa_flags = SA_NOCLDSTOP|SA_NOCLDWAIT|SA_RESTART,
198 log_info("Executing crash shell in 10s...");
201 /* Let the kernel reap children for us */
202 assert_se(sigaction(SIGCHLD, &sa, NULL) == 0);
206 log_error("Failed to fork off crash shell: %m");
208 make_console_stdio();
209 execl("/bin/sh", "/bin/sh", NULL);
211 log_error("execl() failed: %m");
215 log_info("Successfully spawned crash shell as pid "PID_FMT".", pid);
218 log_info("Freezing execution.");
222 static void install_crash_handler(void) {
223 struct sigaction sa = {
225 .sa_flags = SA_NODEFER,
228 sigaction_many(&sa, SIGNALS_CRASH_HANDLER, -1);
231 static int console_setup(bool do_reset) {
234 /* If we are init, we connect stdin/stdout/stderr to /dev/null
235 * and make sure we don't have a controlling tty. */
242 tty_fd = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC);
244 log_error("Failed to open /dev/console: %s", strerror(-tty_fd));
248 /* We don't want to force text mode.
249 * plymouth may be showing pictures already from initrd. */
250 r = reset_terminal_fd(tty_fd, false);
252 log_error("Failed to reset /dev/console: %s", strerror(-r));
258 static int set_default_unit(const char *u) {
267 free(arg_default_unit);
268 arg_default_unit = c;
273 static int parse_proc_cmdline_item(const char *key, const char *value) {
275 static const char * const rlmap[] = {
276 "emergency", SPECIAL_EMERGENCY_TARGET,
277 "-b", SPECIAL_EMERGENCY_TARGET,
278 "single", SPECIAL_RESCUE_TARGET,
279 "-s", SPECIAL_RESCUE_TARGET,
280 "s", SPECIAL_RESCUE_TARGET,
281 "S", SPECIAL_RESCUE_TARGET,
282 "1", SPECIAL_RESCUE_TARGET,
283 "2", SPECIAL_RUNLEVEL2_TARGET,
284 "3", SPECIAL_RUNLEVEL3_TARGET,
285 "4", SPECIAL_RUNLEVEL4_TARGET,
286 "5", SPECIAL_RUNLEVEL5_TARGET,
292 if (streq(key, "systemd.unit") && value) {
295 return set_default_unit(value);
297 } else if (streq(key, "rd.systemd.unit") && value) {
300 return set_default_unit(value);
302 } else if (streq(key, "systemd.log_target") && value) {
304 if (log_set_target_from_string(value) < 0)
305 log_warning("Failed to parse log target %s. Ignoring.", value);
307 } else if (streq(key, "systemd.log_level") && value) {
309 if (log_set_max_level_from_string(value) < 0)
310 log_warning("Failed to parse log level %s. Ignoring.", value);
312 } else if (streq(key, "systemd.log_color") && value) {
314 if (log_show_color_from_string(value) < 0)
315 log_warning("Failed to parse log color setting %s. Ignoring.", value);
317 } else if (streq(key, "systemd.log_location") && value) {
319 if (log_show_location_from_string(value) < 0)
320 log_warning("Failed to parse log location setting %s. Ignoring.", value);
322 } else if (streq(key, "systemd.dump_core") && value) {
324 r = parse_boolean(value);
326 log_warning("Failed to parse dump core switch %s. Ignoring.", value);
330 } else if (streq(key, "systemd.crash_shell") && value) {
332 r = parse_boolean(value);
334 log_warning("Failed to parse crash shell switch %s. Ignoring.", value);
338 } else if (streq(key, "systemd.crash_chvt") && value) {
340 if (safe_atoi(value, &r) < 0)
341 log_warning("Failed to parse crash chvt switch %s. Ignoring.", value);
345 } else if (streq(key, "systemd.confirm_spawn") && value) {
347 r = parse_boolean(value);
349 log_warning("Failed to parse confirm spawn switch %s. Ignoring.", value);
351 arg_confirm_spawn = r;
353 } else if (streq(key, "systemd.show_status") && value) {
355 r = parse_show_status(value, &arg_show_status);
357 log_warning("Failed to parse show status switch %s. Ignoring.", value);
359 } else if (streq(key, "systemd.default_standard_output") && value) {
361 r = exec_output_from_string(value);
363 log_warning("Failed to parse default standard output switch %s. Ignoring.", value);
365 arg_default_std_output = r;
367 } else if (streq(key, "systemd.default_standard_error") && value) {
369 r = exec_output_from_string(value);
371 log_warning("Failed to parse default standard error switch %s. Ignoring.", value);
373 arg_default_std_error = r;
375 } else if (streq(key, "systemd.setenv") && value) {
377 if (env_assignment_is_valid(value)) {
380 env = strv_env_set(arg_default_environment, value);
382 arg_default_environment = env;
384 log_warning("Setting environment variable '%s' failed, ignoring: %s", value, strerror(ENOMEM));
386 log_warning("Environment variable name '%s' is not valid. Ignoring.", value);
388 } else if (streq(key, "quiet") && !value) {
390 log_set_max_level(LOG_NOTICE);
392 if (arg_show_status == _SHOW_STATUS_UNSET)
393 arg_show_status = SHOW_STATUS_AUTO;
395 } else if (streq(key, "debug") && !value) {
397 log_set_max_level(LOG_DEBUG);
399 if (detect_container(NULL) > 0)
400 log_set_target(LOG_TARGET_CONSOLE);
402 } else if (!in_initrd() && !value) {
405 /* SysV compatibility */
406 for (i = 0; i < ELEMENTSOF(rlmap); i += 2)
407 if (streq(key, rlmap[i]))
408 return set_default_unit(rlmap[i+1]);
414 #define DEFINE_SETTER(name, func, descr) \
415 static int name(const char *unit, \
416 const char *filename, \
418 const char *section, \
419 unsigned section_line, \
420 const char *lvalue, \
422 const char *rvalue, \
434 log_syntax(unit, LOG_ERR, filename, line, -r, \
435 "Invalid " descr "'%s': %s", \
436 rvalue, strerror(-r)); \
441 DEFINE_SETTER(config_parse_level2, log_set_max_level_from_string, "log level")
442 DEFINE_SETTER(config_parse_target, log_set_target_from_string, "target")
443 DEFINE_SETTER(config_parse_color, log_show_color_from_string, "color" )
444 DEFINE_SETTER(config_parse_location, log_show_location_from_string, "location")
446 static int config_parse_cpu_affinity2(
448 const char *filename,
451 unsigned section_line,
458 const char *word, *state;
467 FOREACH_WORD_QUOTED(word, l, rvalue, state) {
472 if (!(t = strndup(word, l)))
475 r = safe_atou(t, &cpu);
479 if (!(c = cpu_set_malloc(&ncpus)))
482 if (r < 0 || cpu >= ncpus) {
483 log_syntax(unit, LOG_ERR, filename, line, -r,
484 "Failed to parse CPU affinity '%s'", rvalue);
489 CPU_SET_S(cpu, CPU_ALLOC_SIZE(ncpus), c);
493 if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus), c) < 0)
494 log_warning_unit(unit, "Failed to set CPU affinity: %m");
502 static int config_parse_show_status(
504 const char *filename,
507 unsigned section_line,
515 ShowStatus *b = data;
522 k = parse_show_status(rvalue, b);
524 log_syntax(unit, LOG_ERR, filename, line, -k,
525 "Failed to parse show status setting, ignoring: %s", rvalue);
532 static void strv_free_free(char ***l) {
544 static void free_join_controllers(void) {
545 strv_free_free(arg_join_controllers);
546 arg_join_controllers = NULL;
549 static int config_parse_join_controllers(const char *unit,
550 const char *filename,
553 unsigned section_line,
561 const char *word, *state;
568 free_join_controllers();
570 FOREACH_WORD_QUOTED(word, length, rvalue, state) {
573 s = strndup(word, length);
577 l = strv_split(s, ",");
582 if (strv_length(l) <= 1) {
587 if (!arg_join_controllers) {
588 arg_join_controllers = new(char**, 2);
589 if (!arg_join_controllers) {
594 arg_join_controllers[0] = l;
595 arg_join_controllers[1] = NULL;
602 t = new0(char**, n+2);
610 for (a = arg_join_controllers; *a; a++) {
612 if (strv_overlap(*a, l)) {
613 if (strv_extend_strv(&l, *a) < 0) {
633 t[n++] = strv_uniq(l);
635 strv_free_free(arg_join_controllers);
636 arg_join_controllers = t;
643 static int parse_config_file(void) {
645 const ConfigTableItem items[] = {
646 { "Manager", "LogLevel", config_parse_level2, 0, NULL },
647 { "Manager", "LogTarget", config_parse_target, 0, NULL },
648 { "Manager", "LogColor", config_parse_color, 0, NULL },
649 { "Manager", "LogLocation", config_parse_location, 0, NULL },
650 { "Manager", "DumpCore", config_parse_bool, 0, &arg_dump_core },
651 { "Manager", "CrashShell", config_parse_bool, 0, &arg_crash_shell },
652 { "Manager", "ShowStatus", config_parse_show_status, 0, &arg_show_status },
653 { "Manager", "CrashChVT", config_parse_int, 0, &arg_crash_chvt },
654 { "Manager", "CPUAffinity", config_parse_cpu_affinity2, 0, NULL },
655 { "Manager", "JoinControllers", config_parse_join_controllers, 0, &arg_join_controllers },
656 { "Manager", "RuntimeWatchdogSec", config_parse_sec, 0, &arg_runtime_watchdog },
657 { "Manager", "ShutdownWatchdogSec", config_parse_sec, 0, &arg_shutdown_watchdog },
658 { "Manager", "CapabilityBoundingSet", config_parse_bounding_set, 0, &arg_capability_bounding_set_drop },
660 { "Manager", "SystemCallArchitectures", config_parse_syscall_archs, 0, &arg_syscall_archs },
662 { "Manager", "TimerSlackNSec", config_parse_nsec, 0, &arg_timer_slack_nsec },
663 { "Manager", "DefaultTimerAccuracySec", config_parse_sec, 0, &arg_default_timer_accuracy_usec },
664 { "Manager", "DefaultStandardOutput", config_parse_output, 0, &arg_default_std_output },
665 { "Manager", "DefaultStandardError", config_parse_output, 0, &arg_default_std_error },
666 { "Manager", "DefaultTimeoutStartSec", config_parse_sec, 0, &arg_default_timeout_start_usec },
667 { "Manager", "DefaultTimeoutStopSec", config_parse_sec, 0, &arg_default_timeout_stop_usec },
668 { "Manager", "DefaultRestartSec", config_parse_sec, 0, &arg_default_restart_usec },
669 { "Manager", "DefaultStartLimitInterval", config_parse_sec, 0, &arg_default_start_limit_interval },
670 { "Manager", "DefaultStartLimitBurst", config_parse_unsigned, 0, &arg_default_start_limit_burst },
671 { "Manager", "DefaultEnvironment", config_parse_environ, 0, &arg_default_environment },
672 { "Manager", "DefaultLimitCPU", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_CPU] },
673 { "Manager", "DefaultLimitFSIZE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_FSIZE] },
674 { "Manager", "DefaultLimitDATA", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_DATA] },
675 { "Manager", "DefaultLimitSTACK", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_STACK] },
676 { "Manager", "DefaultLimitCORE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_CORE] },
677 { "Manager", "DefaultLimitRSS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RSS] },
678 { "Manager", "DefaultLimitNOFILE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NOFILE] },
679 { "Manager", "DefaultLimitAS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_AS] },
680 { "Manager", "DefaultLimitNPROC", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NPROC] },
681 { "Manager", "DefaultLimitMEMLOCK", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_MEMLOCK] },
682 { "Manager", "DefaultLimitLOCKS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_LOCKS] },
683 { "Manager", "DefaultLimitSIGPENDING", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_SIGPENDING] },
684 { "Manager", "DefaultLimitMSGQUEUE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_MSGQUEUE] },
685 { "Manager", "DefaultLimitNICE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NICE] },
686 { "Manager", "DefaultLimitRTPRIO", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RTPRIO] },
687 { "Manager", "DefaultLimitRTTIME", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RTTIME] },
688 { "Manager", "DefaultCPUAccounting", config_parse_bool, 0, &arg_default_cpu_accounting },
689 { "Manager", "DefaultBlockIOAccounting", config_parse_bool, 0, &arg_default_blockio_accounting },
690 { "Manager", "DefaultMemoryAccounting", config_parse_bool, 0, &arg_default_memory_accounting },
696 fn = arg_running_as == SYSTEMD_SYSTEM ? PKGSYSCONFDIR "/system.conf" : PKGSYSCONFDIR "/user.conf";
697 config_parse(NULL, fn, NULL,
699 config_item_table_lookup, items,
700 false, false, true, NULL);
705 static int parse_argv(int argc, char *argv[]) {
708 ARG_LOG_LEVEL = 0x100,
718 ARG_DUMP_CONFIGURATION_ITEMS,
725 ARG_DEFAULT_STD_OUTPUT,
726 ARG_DEFAULT_STD_ERROR
729 static const struct option options[] = {
730 { "log-level", required_argument, NULL, ARG_LOG_LEVEL },
731 { "log-target", required_argument, NULL, ARG_LOG_TARGET },
732 { "log-color", optional_argument, NULL, ARG_LOG_COLOR },
733 { "log-location", optional_argument, NULL, ARG_LOG_LOCATION },
734 { "unit", required_argument, NULL, ARG_UNIT },
735 { "system", no_argument, NULL, ARG_SYSTEM },
736 { "user", no_argument, NULL, ARG_USER },
737 { "test", no_argument, NULL, ARG_TEST },
738 { "no-pager", no_argument, NULL, ARG_NO_PAGER },
739 { "help", no_argument, NULL, 'h' },
740 { "version", no_argument, NULL, ARG_VERSION },
741 { "dump-configuration-items", no_argument, NULL, ARG_DUMP_CONFIGURATION_ITEMS },
742 { "dump-core", optional_argument, NULL, ARG_DUMP_CORE },
743 { "crash-shell", optional_argument, NULL, ARG_CRASH_SHELL },
744 { "confirm-spawn", optional_argument, NULL, ARG_CONFIRM_SPAWN },
745 { "show-status", optional_argument, NULL, ARG_SHOW_STATUS },
746 { "deserialize", required_argument, NULL, ARG_DESERIALIZE },
747 { "switched-root", no_argument, NULL, ARG_SWITCHED_ROOT },
748 { "default-standard-output", required_argument, NULL, ARG_DEFAULT_STD_OUTPUT, },
749 { "default-standard-error", required_argument, NULL, ARG_DEFAULT_STD_ERROR, },
761 while ((c = getopt_long(argc, argv, "hDbsz:", options, NULL)) >= 0)
766 r = log_set_max_level_from_string(optarg);
768 log_error("Failed to parse log level %s.", optarg);
775 r = log_set_target_from_string(optarg);
777 log_error("Failed to parse log target %s.", optarg);
786 r = log_show_color_from_string(optarg);
788 log_error("Failed to parse log color setting %s.", optarg);
792 log_show_color(true);
796 case ARG_LOG_LOCATION:
798 r = log_show_location_from_string(optarg);
800 log_error("Failed to parse log location setting %s.", optarg);
804 log_show_location(true);
808 case ARG_DEFAULT_STD_OUTPUT:
809 r = exec_output_from_string(optarg);
811 log_error("Failed to parse default standard output setting %s.", optarg);
814 arg_default_std_output = r;
817 case ARG_DEFAULT_STD_ERROR:
818 r = exec_output_from_string(optarg);
820 log_error("Failed to parse default standard error output setting %s.", optarg);
823 arg_default_std_error = r;
828 r = set_default_unit(optarg);
830 log_error("Failed to set default unit %s: %s", optarg, strerror(-r));
837 arg_running_as = SYSTEMD_SYSTEM;
841 arg_running_as = SYSTEMD_USER;
845 arg_action = ACTION_TEST;
846 if (arg_no_pager < 0)
855 arg_action = ACTION_VERSION;
858 case ARG_DUMP_CONFIGURATION_ITEMS:
859 arg_action = ACTION_DUMP_CONFIGURATION_ITEMS;
863 r = optarg ? parse_boolean(optarg) : 1;
865 log_error("Failed to parse dump core boolean %s.", optarg);
871 case ARG_CRASH_SHELL:
872 r = optarg ? parse_boolean(optarg) : 1;
874 log_error("Failed to parse crash shell boolean %s.", optarg);
880 case ARG_CONFIRM_SPAWN:
881 r = optarg ? parse_boolean(optarg) : 1;
883 log_error("Failed to parse confirm spawn boolean %s.", optarg);
886 arg_confirm_spawn = r;
889 case ARG_SHOW_STATUS:
891 r = parse_show_status(optarg, &arg_show_status);
893 log_error("Failed to parse show status boolean %s.", optarg);
897 arg_show_status = SHOW_STATUS_YES;
900 case ARG_DESERIALIZE: {
904 r = safe_atoi(optarg, &fd);
905 if (r < 0 || fd < 0) {
906 log_error("Failed to parse deserialize option %s.", optarg);
907 return r < 0 ? r : -EINVAL;
910 fd_cloexec(fd, true);
914 log_error("Failed to open serialization fd: %m");
918 if (arg_serialization)
919 fclose(arg_serialization);
921 arg_serialization = f;
926 case ARG_SWITCHED_ROOT:
927 arg_switched_root = true;
931 arg_action = ACTION_HELP;
932 if (arg_no_pager < 0)
937 log_set_max_level(LOG_DEBUG);
943 /* Just to eat away the sysvinit kernel
944 * cmdline args without getopt() error
945 * messages that we'll parse in
946 * parse_proc_cmdline_word() or ignore. */
951 log_error("Unknown option code %c", c);
958 if (optind < argc && getpid() != 1) {
959 /* Hmm, when we aren't run as init system
960 * let's complain about excess arguments */
962 log_error("Excess arguments.");
966 if (detect_container(NULL) > 0) {
969 /* All /proc/cmdline arguments the kernel didn't
970 * understand it passed to us. We're not really
971 * interested in that usually since /proc/cmdline is
972 * more interesting and complete. With one exception:
973 * if we are run in a container /proc/cmdline is not
974 * relevant for the container, hence we rely on argv[]
977 for (a = argv; a < argv + argc; a++) {
978 _cleanup_free_ char *w;
985 value = strchr(w, '=');
989 r = parse_proc_cmdline_item(w, value);
991 log_error("Failed on cmdline argument %s: %s", *a, strerror(-r));
1000 static int help(void) {
1002 printf("%s [OPTIONS...]\n\n"
1003 "Starts up and maintains the system or user services.\n\n"
1004 " -h --help Show this help\n"
1005 " --test Determine startup sequence, dump it and exit\n"
1006 " --no-pager Do not pipe output into a pager\n"
1007 " --dump-configuration-items Dump understood unit configuration items\n"
1008 " --unit=UNIT Set default unit\n"
1009 " --system Run a system instance, even if PID != 1\n"
1010 " --user Run a user instance\n"
1011 " --dump-core[=0|1] Dump core on crash\n"
1012 " --crash-shell[=0|1] Run shell on crash\n"
1013 " --confirm-spawn[=0|1] Ask for confirmation when spawning processes\n"
1014 " --show-status[=0|1] Show status updates on the console during bootup\n"
1015 " --log-target=TARGET Set log target (console, journal, syslog, kmsg, journal-or-kmsg, syslog-or-kmsg, null)\n"
1016 " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
1017 " --log-color[=0|1] Highlight important log messages\n"
1018 " --log-location[=0|1] Include code location in log messages\n"
1019 " --default-standard-output= Set default standard output for services\n"
1020 " --default-standard-error= Set default standard error output for services\n",
1021 program_invocation_short_name);
1026 static int version(void) {
1027 puts(PACKAGE_STRING);
1028 puts(SYSTEMD_FEATURES);
1033 static int prepare_reexecute(Manager *m, FILE **_f, FDSet **_fds, bool switching_root) {
1042 r = manager_open_serialization(m, &f);
1044 log_error("Failed to create serialization file: %s", strerror(-r));
1048 /* Make sure nothing is really destructed when we shut down */
1050 bus_manager_send_reloading(m, true);
1055 log_error("Failed to allocate fd set: %s", strerror(-r));
1059 r = manager_serialize(m, f, fds, switching_root);
1061 log_error("Failed to serialize state: %s", strerror(-r));
1065 if (fseeko(f, 0, SEEK_SET) < 0) {
1066 log_error("Failed to rewind serialization fd: %m");
1070 r = fd_cloexec(fileno(f), false);
1072 log_error("Failed to disable O_CLOEXEC for serialization: %s", strerror(-r));
1076 r = fdset_cloexec(fds, false);
1078 log_error("Failed to disable O_CLOEXEC for serialization fds: %s", strerror(-r));
1096 static int bump_rlimit_nofile(struct rlimit *saved_rlimit) {
1100 assert(saved_rlimit);
1102 /* Save the original RLIMIT_NOFILE so that we can reset it
1103 * later when transitioning from the initrd to the main
1104 * systemd or suchlike. */
1105 if (getrlimit(RLIMIT_NOFILE, saved_rlimit) < 0) {
1106 log_error("Reading RLIMIT_NOFILE failed: %m");
1110 /* Make sure forked processes get the default kernel setting */
1111 if (!arg_default_rlimit[RLIMIT_NOFILE]) {
1114 rl = newdup(struct rlimit, saved_rlimit, 1);
1118 arg_default_rlimit[RLIMIT_NOFILE] = rl;
1121 /* Bump up the resource limit for ourselves substantially */
1122 nl.rlim_cur = nl.rlim_max = 64*1024;
1123 r = setrlimit_closest(RLIMIT_NOFILE, &nl);
1125 log_error("Setting RLIMIT_NOFILE failed: %s", strerror(-r));
1132 static void test_mtab(void) {
1134 static const char ok[] =
1135 "/proc/self/mounts\0"
1137 "../proc/self/mounts\0"
1140 _cleanup_free_ char *p = NULL;
1143 /* Check that /etc/mtab is a symlink to the right place or
1144 * non-existing. But certainly not a file, or a symlink to
1145 * some weird place... */
1147 r = readlink_malloc("/etc/mtab", &p);
1150 if (r >= 0 && nulstr_contains(ok, p))
1153 log_warning("/etc/mtab is not a symlink or not pointing to /proc/self/mounts. "
1154 "This is not supported anymore. "
1155 "Please make sure to replace this file by a symlink to avoid incorrect or misleading mount(8) output.");
1158 static void test_usr(void) {
1160 /* Check that /usr is not a separate fs */
1162 if (dir_is_empty("/usr") <= 0)
1165 log_warning("/usr appears to be on its own filesytem and is not already mounted. This is not a supported setup. "
1166 "Some things will probably break (sometimes even silently) in mysterious ways. "
1167 "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1170 static int initialize_join_controllers(void) {
1171 /* By default, mount "cpu" + "cpuacct" together, and "net_cls"
1172 * + "net_prio". We'd like to add "cpuset" to the mix, but
1173 * "cpuset" does't really work for groups with no initialized
1176 arg_join_controllers = new(char**, 3);
1177 if (!arg_join_controllers)
1180 arg_join_controllers[0] = strv_new("cpu", "cpuacct", NULL);
1181 arg_join_controllers[1] = strv_new("net_cls", "net_prio", NULL);
1182 arg_join_controllers[2] = NULL;
1184 if (!arg_join_controllers[0] || !arg_join_controllers[1]) {
1185 free_join_controllers();
1192 static int enforce_syscall_archs(Set *archs) {
1194 scmp_filter_ctx *seccomp;
1199 seccomp = seccomp_init(SCMP_ACT_ALLOW);
1203 SET_FOREACH(id, arg_syscall_archs, i) {
1204 r = seccomp_arch_add(seccomp, PTR_TO_UINT32(id) - 1);
1208 log_error("Failed to add architecture to seccomp: %s", strerror(-r));
1213 r = seccomp_attr_set(seccomp, SCMP_FLTATR_CTL_NNP, 0);
1215 log_error("Failed to unset NO_NEW_PRIVS: %s", strerror(-r));
1219 r = seccomp_load(seccomp);
1221 log_error("Failed to add install architecture seccomp: %s", strerror(-r));
1224 seccomp_release(seccomp);
1231 static int status_welcome(void) {
1232 _cleanup_free_ char *pretty_name = NULL, *ansi_color = NULL;
1235 r = parse_env_file("/etc/os-release", NEWLINE,
1236 "PRETTY_NAME", &pretty_name,
1237 "ANSI_COLOR", &ansi_color,
1240 r = parse_env_file("/usr/lib/os-release", NEWLINE,
1241 "PRETTY_NAME", &pretty_name,
1242 "ANSI_COLOR", &ansi_color,
1246 if (r < 0 && r != -ENOENT)
1247 log_warning("Failed to read os-release file: %s", strerror(-r));
1249 return status_printf(NULL, false, false,
1250 "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
1251 isempty(ansi_color) ? "1" : ansi_color,
1252 isempty(pretty_name) ? "Linux" : pretty_name);
1255 static int write_container_id(void) {
1258 c = getenv("container");
1262 return write_string_file("/run/systemd/container", c);
1265 int main(int argc, char *argv[]) {
1267 int r, retval = EXIT_FAILURE;
1268 usec_t before_startup, after_startup;
1269 char timespan[FORMAT_TIMESPAN_MAX];
1271 bool reexecute = false;
1272 const char *shutdown_verb = NULL;
1273 dual_timestamp initrd_timestamp = { 0ULL, 0ULL };
1274 dual_timestamp userspace_timestamp = { 0ULL, 0ULL };
1275 dual_timestamp kernel_timestamp = { 0ULL, 0ULL };
1276 dual_timestamp security_start_timestamp = { 0ULL, 0ULL };
1277 dual_timestamp security_finish_timestamp = { 0ULL, 0ULL };
1278 static char systemd[] = "systemd";
1279 bool skip_setup = false;
1281 bool loaded_policy = false;
1282 bool arm_reboot_watchdog = false;
1283 bool queue_default_job = false;
1284 bool empty_etc = false;
1285 char *switch_root_dir = NULL, *switch_root_init = NULL;
1286 static struct rlimit saved_rlimit_nofile = { 0, 0 };
1288 #ifdef HAVE_SYSV_COMPAT
1289 if (getpid() != 1 && strstr(program_invocation_short_name, "init")) {
1290 /* This is compatibility support for SysV, where
1291 * calling init as a user is identical to telinit. */
1294 execv(SYSTEMCTL_BINARY_PATH, argv);
1295 log_error("Failed to exec " SYSTEMCTL_BINARY_PATH ": %m");
1300 dual_timestamp_from_monotonic(&kernel_timestamp, 0);
1301 dual_timestamp_get(&userspace_timestamp);
1303 /* Determine if this is a reexecution or normal bootup. We do
1304 * the full command line parsing much later, so let's just
1305 * have a quick peek here. */
1306 if (strv_find(argv+1, "--deserialize"))
1309 /* If we have switched root, do all the special setup
1311 if (strv_find(argv+1, "--switched-root"))
1314 /* If we get started via the /sbin/init symlink then we are
1315 called 'init'. After a subsequent reexecution we are then
1316 called 'systemd'. That is confusing, hence let's call us
1317 systemd right-away. */
1318 program_invocation_short_name = systemd;
1319 prctl(PR_SET_NAME, systemd);
1324 log_show_color(isatty(STDERR_FILENO) > 0);
1326 /* Disable the umask logic */
1330 if (getpid() == 1 && detect_container(NULL) <= 0) {
1332 /* Running outside of a container as PID 1 */
1333 arg_running_as = SYSTEMD_SYSTEM;
1335 log_set_target(LOG_TARGET_KMSG);
1339 initrd_timestamp = userspace_timestamp;
1342 mount_setup_early();
1343 dual_timestamp_get(&security_start_timestamp);
1344 if (selinux_setup(&loaded_policy) < 0)
1346 if (ima_setup() < 0)
1348 if (smack_setup(&loaded_policy) < 0)
1350 dual_timestamp_get(&security_finish_timestamp);
1353 if (label_init(NULL) < 0)
1357 if (clock_is_localtime() > 0) {
1361 * The very first call of settimeofday() also does a time warp in the kernel.
1363 * In the rtc-in-local time mode, we set the kernel's timezone, and rely on
1364 * external tools to take care of maintaining the RTC and do all adjustments.
1365 * This matches the behavior of Windows, which leaves the RTC alone if the
1366 * registry tells that the RTC runs in UTC.
1368 r = clock_set_timezone(&min);
1370 log_error("Failed to apply local time delta, ignoring: %s", strerror(-r));
1372 log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min);
1373 } else if (!in_initrd()) {
1375 * Do a dummy very first call to seal the kernel's time warp magic.
1377 * Do not call this this from inside the initrd. The initrd might not
1378 * carry /etc/adjtime with LOCAL, but the real system could be set up
1379 * that way. In such case, we need to delay the time-warp or the sealing
1380 * until we reach the real system.
1382 * Do no set the kernel's timezone. The concept of local time cannot
1383 * be supported reliably, the time will jump or be incorrect at every daylight
1384 * saving time change. All kernel local time concepts will be treated
1387 clock_reset_timewarp();
1391 /* Set the default for later on, but don't actually
1392 * open the logs like this for now. Note that if we
1393 * are transitioning from the initrd there might still
1394 * be journal fd open, and we shouldn't attempt
1395 * opening that before we parsed /proc/cmdline which
1396 * might redirect output elsewhere. */
1397 log_set_target(LOG_TARGET_JOURNAL_OR_KMSG);
1399 } else if (getpid() == 1) {
1400 /* Running inside a container, as PID 1 */
1401 arg_running_as = SYSTEMD_SYSTEM;
1402 log_set_target(LOG_TARGET_CONSOLE);
1403 log_close_console(); /* force reopen of /dev/console */
1406 /* For the later on, see above... */
1407 log_set_target(LOG_TARGET_JOURNAL);
1409 /* clear the kernel timestamp,
1410 * because we are in a container */
1411 kernel_timestamp.monotonic = 0ULL;
1412 kernel_timestamp.realtime = 0ULL;
1415 /* Running as user instance */
1416 arg_running_as = SYSTEMD_USER;
1417 log_set_target(LOG_TARGET_AUTO);
1420 /* clear the kernel timestamp,
1421 * because we are not PID 1 */
1422 kernel_timestamp.monotonic = 0ULL;
1423 kernel_timestamp.realtime = 0ULL;
1426 /* Initialize default unit */
1427 r = set_default_unit(SPECIAL_DEFAULT_TARGET);
1429 log_error("Failed to set default unit %s: %s", SPECIAL_DEFAULT_TARGET, strerror(-r));
1433 r = initialize_join_controllers();
1437 /* Mount /proc, /sys and friends, so that /proc/cmdline and
1438 * /proc/$PID/fd is available. */
1439 if (getpid() == 1) {
1440 r = mount_setup(loaded_policy);
1445 /* Reset all signal handlers. */
1446 assert_se(reset_all_signal_handlers() == 0);
1448 ignore_signals(SIGNALS_IGNORE, -1);
1450 if (parse_config_file() < 0)
1453 if (arg_running_as == SYSTEMD_SYSTEM)
1454 if (parse_proc_cmdline(parse_proc_cmdline_item) < 0)
1457 log_parse_environment();
1459 if (parse_argv(argc, argv) < 0)
1462 if (arg_action == ACTION_TEST &&
1464 log_error("Don't run test mode as root.");
1468 if (arg_running_as == SYSTEMD_USER &&
1469 arg_action == ACTION_RUN &&
1471 log_error("Trying to run as user instance, but the system has not been booted with systemd.");
1475 if (arg_running_as == SYSTEMD_SYSTEM &&
1476 arg_action == ACTION_RUN &&
1477 running_in_chroot() > 0) {
1478 log_error("Cannot be run in a chroot() environment.");
1482 if (arg_action == ACTION_TEST)
1485 pager_open_if_enabled();
1487 if (arg_action == ACTION_HELP) {
1490 } else if (arg_action == ACTION_VERSION) {
1493 } else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) {
1494 unit_dump_config_items(stdout);
1495 retval = EXIT_SUCCESS;
1497 } else if (arg_action == ACTION_DONE) {
1498 retval = EXIT_SUCCESS;
1502 if (arg_running_as == SYSTEMD_USER &&
1503 !getenv("XDG_RUNTIME_DIR")) {
1504 log_error("Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
1508 assert_se(arg_action == ACTION_RUN || arg_action == ACTION_TEST);
1510 /* Close logging fds, in order not to confuse fdset below */
1513 /* Remember open file descriptors for later deserialization */
1514 r = fdset_new_fill(&fds);
1516 log_error("Failed to allocate fd set: %s", strerror(-r));
1519 fdset_cloexec(fds, true);
1521 if (arg_serialization)
1522 assert_se(fdset_remove(fds, fileno(arg_serialization)) >= 0);
1524 if (arg_running_as == SYSTEMD_SYSTEM)
1525 /* Become a session leader if we aren't one yet. */
1528 /* Move out of the way, so that we won't block unmounts */
1529 assert_se(chdir("/") == 0);
1531 /* Reset the console, but only if this is really init and we
1532 * are freshly booted */
1533 if (arg_running_as == SYSTEMD_SYSTEM && arg_action == ACTION_RUN)
1534 console_setup(getpid() == 1 && !skip_setup);
1536 /* Open the logging devices, if possible and necessary */
1539 if (arg_show_status == _SHOW_STATUS_UNSET)
1540 arg_show_status = SHOW_STATUS_YES;
1542 /* Make sure we leave a core dump without panicing the
1544 if (getpid() == 1) {
1545 install_crash_handler();
1547 r = mount_cgroup_controllers(arg_join_controllers);
1552 if (arg_running_as == SYSTEMD_SYSTEM) {
1553 const char *virtualization = NULL;
1555 log_info(PACKAGE_STRING " running in %ssystem mode. (" SYSTEMD_FEATURES ")",
1556 arg_action == ACTION_TEST ? "test " : "" );
1558 detect_virtualization(&virtualization);
1560 log_info("Detected virtualization '%s'.", virtualization);
1562 write_container_id();
1564 log_info("Detected architecture '%s'.", architecture_to_string(uname_architecture()));
1567 log_info("Running in initial RAM disk.");
1569 /* Let's check whether /etc is already populated. We
1570 * don't actually really check for that, but use
1571 * /etc/machine-id as flag file. This allows container
1572 * managers and installers to provision a couple of
1573 * files already. If the container manager wants to
1574 * provision the machine ID itself it should pass
1575 * $container_uuid to PID 1.*/
1577 empty_etc = access("/etc/machine-id", F_OK) < 0;
1579 log_info("Running with unpopulated /etc.");
1581 _cleanup_free_ char *t;
1583 t = uid_to_name(getuid());
1584 log_debug(PACKAGE_STRING " running in %suser mode for user "UID_FMT"/%s. (" SYSTEMD_FEATURES ")",
1585 arg_action == ACTION_TEST ? " test" : "", getuid(), t);
1588 if (arg_running_as == SYSTEMD_SYSTEM && !skip_setup) {
1589 if (arg_show_status > 0 || plymouth_running())
1596 machine_id_setup(NULL);
1603 if (arg_running_as == SYSTEMD_SYSTEM && arg_runtime_watchdog > 0)
1604 watchdog_set_timeout(&arg_runtime_watchdog);
1606 if (arg_timer_slack_nsec != NSEC_INFINITY)
1607 if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0)
1608 log_error("Failed to adjust timer slack: %m");
1610 if (arg_capability_bounding_set_drop) {
1611 r = capability_bounding_set_drop_usermode(arg_capability_bounding_set_drop);
1613 log_error("Failed to drop capability bounding set of usermode helpers: %s", strerror(-r));
1616 r = capability_bounding_set_drop(arg_capability_bounding_set_drop, true);
1618 log_error("Failed to drop capability bounding set: %s", strerror(-r));
1623 if (arg_syscall_archs) {
1624 r = enforce_syscall_archs(arg_syscall_archs);
1629 if (arg_running_as == SYSTEMD_USER) {
1630 /* Become reaper of our children */
1631 if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0) {
1632 log_warning("Failed to make us a subreaper: %m");
1633 if (errno == EINVAL)
1634 log_info("Perhaps the kernel version is too old (< 3.4?)");
1638 if (arg_running_as == SYSTEMD_SYSTEM) {
1639 bump_rlimit_nofile(&saved_rlimit_nofile);
1642 r = unit_file_preset_all(UNIT_FILE_SYSTEM, false, NULL, UNIT_FILE_PRESET_FULL, false, NULL, 0);
1644 log_warning("Failed to populate /etc with preset unit settings, ignoring: %s", strerror(-r));
1646 log_info("Populated /etc with preset unit settings.");
1650 r = manager_new(arg_running_as, arg_action == ACTION_TEST, &m);
1652 log_error("Failed to allocate manager object: %s", strerror(-r));
1656 m->confirm_spawn = arg_confirm_spawn;
1657 m->default_timer_accuracy_usec = arg_default_timer_accuracy_usec;
1658 m->default_std_output = arg_default_std_output;
1659 m->default_std_error = arg_default_std_error;
1660 m->default_restart_usec = arg_default_restart_usec;
1661 m->default_timeout_start_usec = arg_default_timeout_start_usec;
1662 m->default_timeout_stop_usec = arg_default_timeout_stop_usec;
1663 m->default_start_limit_interval = arg_default_start_limit_interval;
1664 m->default_start_limit_burst = arg_default_start_limit_burst;
1665 m->default_cpu_accounting = arg_default_cpu_accounting;
1666 m->default_blockio_accounting = arg_default_blockio_accounting;
1667 m->default_memory_accounting = arg_default_memory_accounting;
1668 m->runtime_watchdog = arg_runtime_watchdog;
1669 m->shutdown_watchdog = arg_shutdown_watchdog;
1670 m->userspace_timestamp = userspace_timestamp;
1671 m->kernel_timestamp = kernel_timestamp;
1672 m->initrd_timestamp = initrd_timestamp;
1673 m->security_start_timestamp = security_start_timestamp;
1674 m->security_finish_timestamp = security_finish_timestamp;
1676 manager_set_default_rlimits(m, arg_default_rlimit);
1677 manager_environment_add(m, NULL, arg_default_environment);
1678 manager_set_show_status(m, arg_show_status);
1679 manager_set_first_boot(m, empty_etc);
1681 /* Remember whether we should queue the default job */
1682 queue_default_job = !arg_serialization || arg_switched_root;
1684 before_startup = now(CLOCK_MONOTONIC);
1686 r = manager_startup(m, arg_serialization, fds);
1688 log_error("Failed to fully start up daemon: %s", strerror(-r));
1690 /* This will close all file descriptors that were opened, but
1691 * not claimed by any unit. */
1695 if (arg_serialization) {
1696 fclose(arg_serialization);
1697 arg_serialization = NULL;
1700 if (queue_default_job) {
1701 _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL;
1702 Unit *target = NULL;
1703 Job *default_unit_job;
1705 log_debug("Activating default unit: %s", arg_default_unit);
1707 r = manager_load_unit(m, arg_default_unit, NULL, &error, &target);
1709 log_error("Failed to load default target: %s", bus_error_message(&error, r));
1710 else if (target->load_state == UNIT_ERROR || target->load_state == UNIT_NOT_FOUND)
1711 log_error("Failed to load default target: %s", strerror(-target->load_error));
1712 else if (target->load_state == UNIT_MASKED)
1713 log_error("Default target masked.");
1715 if (!target || target->load_state != UNIT_LOADED) {
1716 log_info("Trying to load rescue target...");
1718 r = manager_load_unit(m, SPECIAL_RESCUE_TARGET, NULL, &error, &target);
1720 log_error("Failed to load rescue target: %s", bus_error_message(&error, r));
1722 } else if (target->load_state == UNIT_ERROR || target->load_state == UNIT_NOT_FOUND) {
1723 log_error("Failed to load rescue target: %s", strerror(-target->load_error));
1725 } else if (target->load_state == UNIT_MASKED) {
1726 log_error("Rescue target masked.");
1731 assert(target->load_state == UNIT_LOADED);
1733 if (arg_action == ACTION_TEST) {
1734 printf("-> By units:\n");
1735 manager_dump_units(m, stdout, "\t");
1738 r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, false, &error, &default_unit_job);
1740 log_debug("Default target could not be isolated, starting instead: %s", bus_error_message(&error, r));
1742 r = manager_add_job(m, JOB_START, target, JOB_REPLACE, false, &error, &default_unit_job);
1744 log_error("Failed to start default target: %s", bus_error_message(&error, r));
1748 log_error("Failed to isolate default target: %s", bus_error_message(&error, r));
1752 m->default_unit_job_id = default_unit_job->id;
1754 after_startup = now(CLOCK_MONOTONIC);
1755 log_full(arg_action == ACTION_TEST ? LOG_INFO : LOG_DEBUG,
1756 "Loaded units and determined initial transaction in %s.",
1757 format_timespan(timespan, sizeof(timespan), after_startup - before_startup, 0));
1759 if (arg_action == ACTION_TEST) {
1760 printf("-> By jobs:\n");
1761 manager_dump_jobs(m, stdout, "\t");
1762 retval = EXIT_SUCCESS;
1768 r = manager_loop(m);
1770 log_error("Failed to run mainloop: %s", strerror(-r));
1774 switch (m->exit_code) {
1777 retval = EXIT_SUCCESS;
1781 case MANAGER_RELOAD:
1782 log_info("Reloading.");
1783 r = manager_reload(m);
1785 log_error("Failed to reload: %s", strerror(-r));
1788 case MANAGER_REEXECUTE:
1790 if (prepare_reexecute(m, &arg_serialization, &fds, false) < 0)
1794 log_notice("Reexecuting.");
1797 case MANAGER_SWITCH_ROOT:
1798 /* Steal the switch root parameters */
1799 switch_root_dir = m->switch_root;
1800 switch_root_init = m->switch_root_init;
1801 m->switch_root = m->switch_root_init = NULL;
1803 if (!switch_root_init)
1804 if (prepare_reexecute(m, &arg_serialization, &fds, true) < 0)
1808 log_notice("Switching root.");
1811 case MANAGER_REBOOT:
1812 case MANAGER_POWEROFF:
1814 case MANAGER_KEXEC: {
1815 static const char * const table[_MANAGER_EXIT_CODE_MAX] = {
1816 [MANAGER_REBOOT] = "reboot",
1817 [MANAGER_POWEROFF] = "poweroff",
1818 [MANAGER_HALT] = "halt",
1819 [MANAGER_KEXEC] = "kexec"
1822 assert_se(shutdown_verb = table[m->exit_code]);
1823 arm_reboot_watchdog = m->exit_code == MANAGER_REBOOT;
1825 log_notice("Shutting down.");
1830 assert_not_reached("Unknown exit code.");
1842 for (j = 0; j < ELEMENTSOF(arg_default_rlimit); j++) {
1843 free(arg_default_rlimit[j]);
1844 arg_default_rlimit[j] = NULL;
1847 free(arg_default_unit);
1848 arg_default_unit = NULL;
1850 free_join_controllers();
1852 strv_free(arg_default_environment);
1853 arg_default_environment = NULL;
1855 set_free(arg_syscall_archs);
1856 arg_syscall_archs = NULL;
1862 unsigned i, args_size;
1865 /* Close and disarm the watchdog, so that the new
1866 * instance can reinitialize it, but doesn't get
1867 * rebooted while we do that */
1868 watchdog_close(true);
1870 /* Reset the RLIMIT_NOFILE to the kernel default, so
1871 * that the new systemd can pass the kernel default to
1872 * its child processes */
1873 if (saved_rlimit_nofile.rlim_cur > 0)
1874 setrlimit(RLIMIT_NOFILE, &saved_rlimit_nofile);
1876 if (switch_root_dir) {
1877 /* Kill all remaining processes from the
1878 * initrd, but don't wait for them, so that we
1879 * can handle the SIGCHLD for them after
1881 broadcast_signal(SIGTERM, false, true);
1883 /* And switch root */
1884 r = switch_root(switch_root_dir);
1886 log_error("Failed to switch root, ignoring: %s", strerror(-r));
1889 args_size = MAX(6, argc+1);
1890 args = newa(const char*, args_size);
1892 if (!switch_root_init) {
1895 /* First try to spawn ourselves with the right
1896 * path, and with full serialization. We do
1897 * this only if the user didn't specify an
1898 * explicit init to spawn. */
1900 assert(arg_serialization);
1903 snprintf(sfd, sizeof(sfd), "%i", fileno(arg_serialization));
1907 args[i++] = SYSTEMD_BINARY_PATH;
1908 if (switch_root_dir)
1909 args[i++] = "--switched-root";
1910 args[i++] = arg_running_as == SYSTEMD_SYSTEM ? "--system" : "--user";
1911 args[i++] = "--deserialize";
1915 /* do not pass along the environment we inherit from the kernel or initrd */
1916 if (switch_root_dir)
1919 assert(i <= args_size);
1920 execv(args[0], (char* const*) args);
1923 /* Try the fallback, if there is any, without any
1924 * serialization. We pass the original argv[] and
1925 * envp[]. (Well, modulo the ordering changes due to
1926 * getopt() in argv[], and some cleanups in envp[],
1927 * but let's hope that doesn't matter.) */
1929 if (arg_serialization) {
1930 fclose(arg_serialization);
1931 arg_serialization = NULL;
1939 /* Reopen the console */
1940 make_console_stdio();
1942 for (j = 1, i = 1; j < (unsigned) argc; j++)
1943 args[i++] = argv[j];
1945 assert(i <= args_size);
1947 /* reenable any blocked signals, especially important
1948 * if we switch from initial ramdisk to init=... */
1949 reset_all_signal_handlers();
1951 assert_se(sigemptyset(&ss) == 0);
1952 assert_se(sigprocmask(SIG_SETMASK, &ss, NULL) == 0);
1954 if (switch_root_init) {
1955 args[0] = switch_root_init;
1956 execv(args[0], (char* const*) args);
1957 log_warning("Failed to execute configured init, trying fallback: %m");
1960 args[0] = "/sbin/init";
1961 execv(args[0], (char* const*) args);
1963 if (errno == ENOENT) {
1964 log_warning("No /sbin/init, trying fallback");
1966 args[0] = "/bin/sh";
1968 execv(args[0], (char* const*) args);
1969 log_error("Failed to execute /bin/sh, giving up: %m");
1971 log_warning("Failed to execute /sbin/init, giving up: %m");
1974 if (arg_serialization) {
1975 fclose(arg_serialization);
1976 arg_serialization = NULL;
1984 #ifdef HAVE_VALGRIND_VALGRIND_H
1985 /* If we are PID 1 and running under valgrind, then let's exit
1986 * here explicitly. valgrind will only generate nice output on
1987 * exit(), not on exec(), hence let's do the former not the
1989 if (getpid() == 1 && RUNNING_ON_VALGRIND)
1993 if (shutdown_verb) {
1994 char log_level[DECIMAL_STR_MAX(int) + 1];
1995 const char* command_line[9] = {
1996 SYSTEMD_SHUTDOWN_BINARY_PATH,
1998 "--log-level", log_level,
2002 _cleanup_strv_free_ char **env_block = NULL;
2004 assert(command_line[pos] == NULL);
2005 env_block = strv_copy(environ);
2007 snprintf(log_level, sizeof(log_level), "%d", log_get_max_level());
2009 switch (log_get_target()) {
2010 case LOG_TARGET_KMSG:
2011 case LOG_TARGET_JOURNAL_OR_KMSG:
2012 case LOG_TARGET_SYSLOG_OR_KMSG:
2013 command_line[pos++] = "kmsg";
2016 case LOG_TARGET_CONSOLE:
2018 command_line[pos++] = "console";
2022 if (log_get_show_color())
2023 command_line[pos++] = "--log-color";
2025 if (log_get_show_location())
2026 command_line[pos++] = "--log-location";
2028 assert(pos < ELEMENTSOF(command_line));
2030 if (arm_reboot_watchdog && arg_shutdown_watchdog > 0) {
2033 /* If we reboot let's set the shutdown
2034 * watchdog and tell the shutdown binary to
2035 * repeatedly ping it */
2036 watchdog_set_timeout(&arg_shutdown_watchdog);
2037 watchdog_close(false);
2039 /* Tell the binary how often to ping, ignore failure */
2040 if (asprintf(&e, "WATCHDOG_USEC="USEC_FMT, arg_shutdown_watchdog) > 0)
2041 strv_push(&env_block, e);
2043 watchdog_close(true);
2045 /* Avoid the creation of new processes forked by the
2046 * kernel; at this point, we will not listen to the
2048 if (detect_container(NULL) <= 0)
2049 cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER);
2051 execve(SYSTEMD_SHUTDOWN_BINARY_PATH, (char **) command_line, env_block);
2052 log_error("Failed to execute shutdown binary, %s: %m",
2053 getpid() == 1 ? "freezing" : "quitting");